基于静电信号的人体动作识别

Human Motion Recognition Based on Electrostatic Signals

提出一种通过检测人体行为动作产生的静电信号进行人体动作识别的方法.在分析人体荷电特性的基础上,设计静电信号检测系统采集被测人员的5种典型动作(行走、踏步、坐下、拿取物品、挥手)的静电感应信号.对采集的5种动作的静电信号进行特征参量提取和显著性差异分析,优化用于分类的特征参数.基于Weka平台使用3种分类算法(支持向量机、决策树C4.5和随机森林)分别对采集到的250组样本数据通过10折交叉验证进行了分类识别,结果显示随机森林算法的识别效果最好,正确率可达99.6%.研究表明本文提出的单人环境下基于人体静电信号的动作分类识别方法能够有效地对典型人体动作进行识别.

A human motion recognition method by detecting electrostatic signals generated by human behaviors is proposed. Based on the analysis of the charge characteristics of human body, a static electricity detection system is designed to collect the electrostatic induction signals of 5 typical actions of the tested persons, i.e. walking, stepping, sitting down,taking the goods, and waving hand. The characteristic parameters of the collected 5 kinds of human body electrostatic signals are extracted, their significant differences are analyzed, and the characteristic parameters for classification are optimized. 3 kinds of classification algorithms including support vector machine, decision tree-C4.5 and random forest, are used based on Weka platform to classify the 250 collected signal samples by 10-fold cross-validation. The results show that the random forest algorithm obtains the best recognition effect with the accuracy of 99.6%. The research shows that the proposed action classification method based on human electrostatic signals for single environment can effectively identify typical human actions.